Upload
prateekkumar
View
217
Download
0
Embed Size (px)
Citation preview
7/26/2019 Simulation in Opnet
1/8
A comparative study of wireless communication protocols for
monitoring vital signs in athletes in a soccer field
G.KIOKES1,P.GKONIS
2,E.ZOUNTOURIDOU
2
1Department of Electronics, Electric Power and Telecommunications,2School of Electrical and Computer Engineering,
1Hellenic Air-Force Academy,
2National Technical University of Athens,1Air Base Dekeleia,
29 Iroon Politechniou str. ,Zografou, Athens
GREECE
[email protected],[email protected], [email protected]
Abstract: - This paper provides a comprehensive investigation of two wireless protocols performance over short
range communications for monitoring vital signs in athletes, mainly during training. Live physiologicalmonitoring of athletes during sports can help to optimize their performance at a specific time in order tomaximize athlete efficiency and preventing undesirable events like injuries. ZigBee (over IEEE 802.15.4) isbecoming a popular way to create wireless personal area network due to its low power consumption andscalability while Wi-Fi (over IEEE 802.11g) provides the solution for portability with connection of mobility aswell. This article presents the evaluation results of system performance using OPNET simulation for the twodifferent wireless standards. The evaluation is concentrated on providing results regarding network values such
as end-to-end delay, traffic, average throughput captured from global and objects statistics.
Key-Words: -Wireless sensor networks, Performance evaluation, OPNET, Simulation, Throughput
1 IntroductionThe evolution on sensors technology over the lastdecades has opened completely new fields for the
modern technological applications. Nowadays,wireless sensors networks (WSNs) [1] constitute apromising field of research in the area of WirelessCommunications, given the extensive breadth ofapplications that they can support. They represent a
new order of operating calculating systems thatextends the interaction between humans and the
natural environment. While up to today the wired
sensors networks were globally used, the growth ofmicroelectronic systems technology in the wirelesscommunications sector, made feasible the use
wireless sensors. The wireless technology providesmore flexibility, easiness of use and fast growth ofsensors networks; although in certain critical
applications there are data transfer reliability issues.More analytically, the WSNs are constituted by oneor more sink or base stations and by tens orthousands sensor nodes, which are scattered in a
space. These nodes collect information from theenvironment and depending on the application,
either process the information and send it to acentral node, or they send it without processing.
These nodes, usually sensor the temperature, thelight, the vibration, the sound, e.t.c.. This
information travels into the network, having asfinal destination the nodes sink. Depending on theapplication, sinks sent queries to the nodes, in orderto gather useful information.The most important advantage of this type ofnetwork is that, the beforehand knowledge of thetopology is not required. This ability allows the
rapid growth of networks in remotely and wildregions. The nodes have low cost and lowconsumption, as they have the ability ofcommunicating in small distances, executing limited
local data process and sensor signals of varioustypes, in their application region. A sensor network
is characterized from its lifetime, its expandability,its coverage, the production cost, the fault detection
and correction, its synchronization, the responsetime, as well as the safety it provides. The storage
energy capacity of the devices is the restrictivefactor of life duration. It should be noted that thecritical characteristic in a lot of applications, is not
the average lifetime of one node, but the minimum
estimated lifetime.Training is a repeating (rollover) process consistingof four steps: assessment, planning, implementation,
WSEAS TRANSACTIONS on COMMUNICATIONS G. Kiokes, P. Gkonis, E. Zountouridou
E-ISSN: 2224-2864 433 Volume 14, 2015
mailto:[email protected]:[email protected]7/26/2019 Simulation in Opnet
2/8
and monitoring [2]. Overtraining is a non desirableprocess of excessive exercise training in high-performance athletes that may lead to overtraining
syndrome. According author [3], overtrainingsyndrome is a neuroendocrine disorder
characterized by poor performance in competition,inability to maintain training loads, persistentfatigue, reduced catecholamine excretion, frequentillness, disturbed sleep and alterations in mood state.
In order to detect overtraining, the training load ofeach athlete needs to be monitored and
individualized. Monitoring training load requiresquantification of the intensity and duration of thephysiological stress imposed on the athlete.Consequently, training volume and intensity are thebasic training variables that characterize training
load [4],[5].
Athlete monitoring is key to achieving both shortterm and long term competitive success. The moreconsistent the monitoring, the more meaningful theinformation will be. The increased athleticcompetiveness and at the same time the technology
revolution, led to discover ways of monitoring theathletes training and performance. Monitoring eachathlete data, such as heart rate, nutrition, providesuseful information, enabling the personalization of
the training, avoiding overtraining, which eventuallylead performance maximization. In recent years,several wireless access technologies are broadening
and changing our habits. Each wireless solutionoffers a specific mix of transmission band, costs,and coverage according to the needs that originatedit.
A comparative study of the two technologies isgiven in [6]. The authors presented a broad
overview of the four most popular wirelessstandards, Bluetooth, UWB, ZigBee, and Wi-Fi witha quantitative valuation in terms of the transmissiontime, data coding efficiency, protocol complexity,and power consumption. The performance analysis
of the 802.15.4 standard has been extensively
studied in the literature such as [7], [8], [9]. Forinstance, in [7] authors investigate the performanceof the aforementioned network through simulationsin the OPNET Modeler environment and came tothe conclusion that the ZigBee protocol is well
suited for monitoring athletes subscribers andsimilar applications.
In this paper, firstly an overview of the mentionedwireless protocols is presented. Following a
performance evaluation comparison of the twostandards in an soccer athletic field area isconducted using OPNET simulator, in order to
investigate the more appropriate protocol formonitoring athletes vital signs by training. A
specific scenario is analyzed and simulated and thennetwork values such as end-to-end delay, traffic andthroughput are exported. The rest of this paper is
organized as follows. Tthe architecture of theprotocols specified for wireless sensor networks is
described in Sec. 2. Following that, Sec. 3 presentsthe simulation results achieved by using OPNETSimulator for Physical and MAC layer performanceestimation. Finally, conclusions are given, offering
quantitative evaluation of the proposed approach inSec. 4.
2 Wireless Technologies verview
2.1 Zigbee
IEEE 802.15.4 standard [10], determines thespecifications of the Physical layer and the MediumAccess Control (MAC) sub-layer of wirelesspersonal area networks (WPANs) using devices theyconsume low power. ZigBee systems are beingdesigned to provide wireless short-range
communications (up to 100m), depending on poweroutput and environmental characteristics.
WSN have particularly different requirements thanstandard local networks of ultra-high datatransmission speed. It is reminded that devices likesensors are able to record and transmit information
to environments inaccessible to human factors forparticularly long periods. This implies that thedesign of each protocol related to them must take
account of the following features:a) the need for low power and energy consumption,
b) the need for self-configuration of the wirelessnetwork and adaptation thereof to the environment
data, andc) the signal shielding mechanism againstinterference, at physical level.
On another aspect, the high transmission speeds
required in WLAN is of no particular concern, asthe information usually transmitted relate to the
recording of certain environment data (that is a fewbytes of information) and not large volume files, asin the case of WLAN.
ZigBee specification focuses only on upperlayers (Fig. 1): starting from the network layer to the
final application layer, including the applicationobjects themselves. The ZigBee application layer
consists of the Application Support sub-layer (APS),the ZigBee Device Object (ZDO) and themanufacturer-defined application objects. The
responsibilities of the APS sub-layer includemaintaining tables for binding, which is the ability
WSEAS TRANSACTIONS on COMMUNICATIONS G. Kiokes, P. Gkonis, E. Zountouridou
E-ISSN: 2224-2864 434 Volume 14, 2015
7/26/2019 Simulation in Opnet
3/8
to match two devices together based on theirservices and their needs, and forwarding messagesbetween bound devices. Another responsibility of
the APS sub-layer is discovery, which is the abilityto determine which other devices are operating in
the personal operating space of a device [11].Wireless connections under 802.15.4 standard
can operate in three ISM (Industrial ScientificMedical) frequency bands with the following data
rates:1) 250kbps in the 2.4 GHz band (O-QPSK
encoding),2) 40kbps in the 915 MHz band (BPSK encoding)and3) 20 kbps in the 868 MHz band (BPSK encoding)
Fig. 1 ZigBee Communication Stack
2.2 The WI-FI standardWireless Local Area Networks (WLANs) are one ofthe internet-related technologies that during the last
decade marked the greatest expansion at globallevel. The replacement of analog lines with digital
ones, the expansion of optics communications (fiberoptics), and the increased information transmission
rates occurred in parallel to the replacement ofcertain wired functions by wireless ones. Currently,
every household, or work area with internet accesssupports a wireless network through the routerantenna and the wireless network card of a laptop.Therefore, it is obvious that any design at least in
an urban environment - of a network other thanWLAN, such as those of 802.15.4 which operate in
the same frequencies and at a lower transmissionstrength, must take into account the interferencecaused by wireless local area networks.
IEEE 802.11 protocol defines two types ofequipment (Fig 2), a wireless station, which is
usually a PC equipped with a wireless networkinterface card and an access point acting as a bridgebetween the wireless and wired areas of the
network. The access point usually consists of anantenna, a wired network interface and a software
that responds to protocol 802.11d, which transfersdata from the wireless to the wired network. Accesspoint operates as a base-station for the wirelessnetwork, allowing multiple terminal devices
connection to the wired network. IEEE 802.11standard defines two operating modes: (a)
infrastructure mode and (b) ad hoc mode [12]
Fig. 2 Wi-Fi network topology
Wi-Fi standard defines the following networkingtopologies:
IBSS (Independent Basic Service Set) or ad hoc
topologyThis is the most simple wireless network topology(Fig. 3). The stations are peers and communicate
directly between them, peer to peer, without acentral station. This networking mode is usedprimarily by small networks. A necessaryprecondition for the communication between two
stations is that each one is within the range of theother.
Fig. 3 IBSS topology
Infrastructure
BSS topology is a more complex networkingtopology in which the wireless network has clusterform. In each cluster there is a number of wireless
WSEAS TRANSACTIONS on COMMUNICATIONS G. Kiokes, P. Gkonis, E. Zountouridou
E-ISSN: 2224-2864 435 Volume 14, 2015
7/26/2019 Simulation in Opnet
4/8
stations (clients), which communicate between themthrough a central distributor BS (Base Station) or,more commonly, AP (Access Point).
There is determination of two service types,depending on the number of Aps, and therefore the
number of clusters [13]:a) Infrastructure Basic Service Set.b) ESS, Extended Service Set
A wireless network can also consist of more than
one clusters, which can be bridged over aDistribution System. The distribution system is the
network connecting APs between them and with theother networks. The entire interconnected network,including the distribution system and the APsconstitutes the ESS topology. That is, it is a set ofclusters, each one of which consists of an access
point AP, with all APs being connected to a network
structure. This achieves larger wireless coveragerange. In this case, each wireless station cancommute from on cluster to the other without thestation itself, or the network experiencing anychange. This ability of the network is called
roaming.IEEE 802.11g protocol operates on the same
bandwidth with 802.11b, namely 22MHz, whichallows it to achieve higher transmission speeds.
Moreover, depending on the environmental noise,the protocol is able to select among new operationspeeds which are 6,9,12,18,24,46,48,54Mbps. In
addition, it is compatible with 802.11b as it can alsosupport 1, 2, 5.5 and 11Mbps speeds using DSSSand CCK encoding. The need to increasetransmission speeds while maintaining bandwidth at
22MHz, in conjunction with the resistance of thesystem to intersymbol interference, were the key
parameters for the use of a new configuration,OFDM (Orthogonal Frequency DivisionMultiplexing). New speeds (6,9,12...54) are theresult of OFDM configuration. OFDM is the soleconfiguration able to achieve high transmission rates
while resisting interference. The following table
depicts the frequency bands and the theoretical andactual speeds of the various 802.11 standards [14]
Table 1 The frequency bands of 802.11 standard
Standard
Frequency Band
Typical
Transmission Rate
Theoretical
Transmission Rate
Range
Indoors
802.11a 5GHz 23Mbps 54Mbps 35m
802.11b 2.4GHz 4.3Mbps 11Mbps 38m
802.11g 2.4GHz 19Mbps 54Mbps 38m
3 Simulation Results
3.1 Simulation model descriptionOPNET Modeler [15] constitutes a specialized tool
in the area of communications, having graphicenvironment for modelling and simulation, ofvarious types of networks. The platform allows the
design and thorough study of telecommunicationnetworks offering great flexibility depending on thelayer of the network concerned. Moreover
simulating separate models for each network, it ispossible to take measurements of global and objectparameters, such as packet delay, packet loss,throughput, e.t.c. and through them can export
general conclusions of the systems performance.The main objective was the development and
study of the two communication networks betweenathletes and coordinator (coach). The area selected
for the simulation was a typical soccer field. Thelength of a full-size soccer pitch must be between
100 yards (90 metres) and 130 yards (120 metres)and the width between 50 yards (45 metres) and 100yards (90 metres) with 11 players. Each of the 11players was able to move in a specific territoryclockwise in a parallelogram trajectory round the
area depending on its playing position. For example,a Goalkeeper player, can move along the sidelines
of the playing area near the goalposts (Fig. 4 andFig. 5). The subscribers are moving with differentspeeds (10 to 15 Km/h). The scenario primaryobjectives were the recording of the communication
link between the nodes and the coach, the collectionof the identification and status data transmitted overa short period of time for each network topologywhereas depending on the results, conclusionswould arise regarding each protocols performance.
Fig. 4. Geographic positioning in simulation environmentfor WLAN IEEE 802.11g protocol
WSEAS TRANSACTIONS on COMMUNICATIONS G. Kiokes, P. Gkonis, E. Zountouridou
E-ISSN: 2224-2864 436 Volume 14, 2015
7/26/2019 Simulation in Opnet
5/8
Fig. 5 Geographic positioning in simulation
environment for ZigBEE protocol.
At the beginning, mobile stations of the network
were requested to send a large amount ofinformation related to their status. For this reason, inthe simulation model, the size of packets transferredvia Database Access application was set to MediumLoad. The duration of the simulation was selected tobe 10 minute due to the amount of data required tobe processed. Data rate was set to 250 kbits/secondin the 2.4 GHz for ZigBee and 54 Mbits/second in
the same frequency band for 802.11g.
3.2. ResultsOne can collect values from individual nodes in thenetwork (node statistics) or from the entire network
(global statistics). Global statistics can be used togather information about the network as a whole.
Node statistics provide information about individualnode such as coordinator, router or end device. The
eleven nodes in the athletic field topology wereidentical. The Global Statistics of the network,whose graphs will be examined, are listed and
analyzed in the following:
End to End delay (sec): is the time that
lapsed between creation and reception of anapplication packet.
Throughput (bits/sec): Represents the totalnumber of bits (in bits/sec) forwarded from wirelessLAN to higher layers in all WLAN nodes of thenetwork.
The Node Statistics for the Coordinator and onenode (player 5) of the network are analyzed below:
Throughput (bits/sec): represents the totaldata traffic in bits/sec successfully received and
forwarded to the higher layer by the WLAN.
Data traffic received (bits/sec): represents
the average number of bitsper second for trafficsuccessfully received by the MAC from the physical
layer. This includes retransmissions.
Data traffic sent (bits/sec):represents the
traffic transmitted by the MAC in bits/sec. While
computing the size of the transmitted packets for
this statistic, the physical layer and MAC headers of
the packet are also included.
The results and graphs are presented in thefollowing Figs 6-10. Fig. 6 demonstrate the end to
end delay of the network for the 802.11g (bluecurve) and ZigBee (red curve) technologies
respectively. As a result, the delay is higher inZigBee than in 802.11g. The main reason for this isthe different data rate according the usedtechnology. The next graph (Fig. 7) represents
network throughput (bps). It can be seen thatthroughput is heavily influenced by the traffic load
since none of the networks achieves theoreticaltransmission values and 802.11g performs better.
Similarly results can be obtain for Fig. 8. It can beenobserved that WIFI throughput is higher andcoordinator does not accept packets until the
pending data are not completely transmittedavoiding collisions situations. Fig. 9,10 represents
traffic received and sent (bits/sec) values forcoordinator and node 5 player. In both cases, it is
observed that WLAN technology continues toperform better. Fig. 11 and 12 also shows that all
nodes have more or less equivalent traffics.Takingeverything into account 802.11g provides fastnetworking connections and outperform the otherstandard.
4. CONCLUSIONSThe objective of this paper is to examine the
performance of the IEEE 802.11g and IEEE802.15.4 standards for wireless sensor networksapplications. At the beginning an overview of thetwo wireless technologies adapted in monitoring
vital signs of athletes, mainly during training ispresented. Furthermore through simulations in
OPNET Modeler environment global and nodestatistics in soccer field area with eleven nodes isexamined. The trajectory on each mobile node isconfigured on specific destination round the field.According to simulation results the 802.11g
protocol tends to outperform IEEE 802.15.4protocol mainly due to its high transmission rate.
More specific it gives less network end to end delayand higher average throughput.
In the future authors intend to perform furtherexperiments in order to examine power consumption
and energy efficiency of the two technologies.
WSEAS TRANSACTIONS on COMMUNICATIONS G. Kiokes, P. Gkonis, E. Zountouridou
E-ISSN: 2224-2864 437 Volume 14, 2015
7/26/2019 Simulation in Opnet
6/8
ZigBee's lower data rates, can support much lowerpower consumption rates. Extending this workauthors will develop the whole physical layer
transmitter and receiver for the two protocols usingNallatech XtremeDSP Development Kit with Xilinx
FPGA for real time implementations and study thetheoretical model variation.
Fig. 6. Global statistics - end to end delay (sec) comparison for both protocols
Fig. 7. Global statisticsThroughput (bits/sec) comparison for both protocols
Object (global) - IEEE 802.15.4(bits/sec) Object (global) - IEEE 802.11g (bits/sec)
Object (global) - IEEE 802.11g(bits/sec) Object (global) - IEEE 802.15.4 (bits/sec)
WSEAS TRANSACTIONS on COMMUNICATIONS G. Kiokes, P. Gkonis, E. Zountouridou
E-ISSN: 2224-2864 438 Volume 14, 2015
7/26/2019 Simulation in Opnet
7/8
Fig. 8. Coordinator statisticsThroughput (bits/sec) comparison for both protocols
Fig. 9. Coordinator statistics -Traffic received, Traffic sent (bits/sec) comparison for both protocols
Object (Coordinator) - IEEE 802.11gTraffic Rcvd (bits/sec) Object (Coordinator) - IEEE 802.15.4 Traffic Rcvd (bits/sec)
Object (Coordinator) - IEEE 802.11g- Traffic Sent (bits/sec) Object (Coordinator) - IEEE 802.15.4 Traffic Sent (bits/sec)
Object (Coordinator) - IEEE 802.15.4(bits/sec) Object (Coordinator) - IEEE 802.11g (bits/sec)
WSEAS TRANSACTIONS on COMMUNICATIONS G. Kiokes, P. Gkonis, E. Zountouridou
E-ISSN: 2224-2864 439 Volume 14, 2015
7/26/2019 Simulation in Opnet
8/8
Fig. 10. Player 5Router device statistics - Traffic received, Traffic sent (bits/sec) comparison for both
protocols
References:
[1] Akyildiz F. and Kasimoglu I.H., Ad HocNetworks 2(4):351-367, 2004
[2] Cambetta V. (2007), Monitoring training iscritical for success.http:// www.Humankinetics .com /excerpts/excerpts/monitoring-training-is-critical-for-
success[3] MacKinnon L. Overtraining effects on
immunity and performance in athletes,Immunology and Cell Biology 78:502509,2000
[4] Monoem H., Behm D., Tebben M. and K.Chamari,Monitoring Training Load, Recovery,
Overtraining and Upper respiratory Infectionin Taekwondo. Performance Optimization inTaekwondo: From Laboratory to Field. OMICSGroup International, 2014
[5] Foster C., Hector LL., Welsh R., Schrager M.,
Green MA., Effects of specific versus cross-training on running performance., Eur J ApplPhysiol Occup Physiol. 70:367-372,1995
[6] Jin-Shyan Lee, Yu-Wei Su, and Chung-ChouShen., A Comparative Study of Wireless
Protocols: Bluetooth, UWB, ZigBee, and Wi-Fi., IEEE Ind. Electr. Society Conf., Taipei,Taiwan, 2007
[7] Kiokes G., Vossou C., Chatzistamatis P., et.all.Performance Evaluation of a CommunicationProtocol for Vital Signs Sensors Used for theMonitoring of Athletes, Int. J. Of Distrib.
Sensor Networks. 2014
[8] Mahajan R., Nair S , Performance Evaluation
of Zigbee Protocol Using Opnet Modeler forMine Safety, Int. J. of Comp. Science and
Network, 2013[9] Mihajlov B. and Bogdanoski M., Overview and
Analysis of the Performances of ZigBee basedWireless Sensor Networks, Int. J. of Comp.
Applications, 2011[10]Feng Xia, Ruonan Hao, Yang Cao, and Lei
Xue, A Survey of Adaptive and Real-TimeProtocols Based on IEEE 802.15.4, Int.J.ofDistrib. Sensor Networks, 2011
[11]OPEN-Meter project ,Description of state- of-the-art wireless access technology, OPEN-
Meter WP2 D2.1 part3 v1.0. 2009[12]Ian Poolehttp://www.radioelectronics.com/info
/wireless/wi-fi/ieee-802-11-standards-tutorial.php Accessed 03 Jan 2015.
[13] Shooshtary S. ,Development of a MATLAB
Simulation Environment for Vehicle-to-Vehicleand Infrastructure Communication Based onIEEE 802.11pMaster Thesis, University ofGavle, Vienna, 2008
[14]IEEE-SA.IEEE 802.11: Wireless LAN Medium
Access Control (MAC) and Physical Layer(PHY) Specifications,2012 revision.
[15] Riverbed (2012), Riverbed application andnetwork performance management solutions.http://www.riverbed.com/products-solutions/products/opnet.html?redirect=opnet
Object (Player 5) - IEEE 802.15.4Traffic Rcvd (bits/sec) Object (Player 5) - IEEE 802.11g Traffic Rcvd (bits/sec)
Object (Player 5) - IEEE 802.15.4- Traffic Sent (bits/sec) Object (Player 5) - IEEE 802.11g Traffic Sent (bits/sec)
WSEAS TRANSACTIONS on COMMUNICATIONS G. Kiokes, P. Gkonis, E. Zountouridou
E-ISSN: 2224-2864 440 Volume 14, 2015
http://www.ncbi.nlm.nih.gov/pubmed/7649149http://www.ncbi.nlm.nih.gov/pubmed/7649149http://www.ncbi.nlm.nih.gov/pubmed/7649149http://www.ncbi.nlm.nih.gov/pubmed/7649149http://www.ncbi.nlm.nih.gov/pubmed/7649149http://localhost/var/www/apps/conversion/tmp/scratch_2/wseas/Ian%20Poolehttp://www.radioelectronics.com/info%20%20/wireless/wi-fi/ieee-802-11-standards-tutorial.%20php%20Accessed%2003%20Jan%202015http://www.radioelectronics.com/info%20%20/wireless/wi-fi/ieee-802-11-standards-tutorial.%20php%20Accessed%2003%20Jan%202015http://www.radioelectronics.com/info%20%20/wireless/wi-fi/ieee-802-11-standards-tutorial.%20php%20Accessed%2003%20Jan%202015http://en.wikipedia.org/wiki/IEEE-SAhttp://standards.ieee.org/getieee802/download/802.11-2012.pdfhttp://standards.ieee.org/getieee802/download/802.11-2012.pdfhttp://standards.ieee.org/getieee802/download/802.11-2012.pdfhttp://standards.ieee.org/getieee802/download/802.11-2012.pdfhttp://www.riverbed.com/products-solutions/products/opnet.html?redirect=opnethttp://www.riverbed.com/products-solutions/products/opnet.html?redirect=opnethttp://www.riverbed.com/products-solutions/products/opnet.html?redirect=opnethttp://www.riverbed.com/products-solutions/products/opnet.html?redirect=opnethttp://standards.ieee.org/getieee802/download/802.11-2012.pdfhttp://standards.ieee.org/getieee802/download/802.11-2012.pdfhttp://standards.ieee.org/getieee802/download/802.11-2012.pdfhttp://en.wikipedia.org/wiki/IEEE-SAhttp://www.radioelectronics.com/info%20%20/wireless/wi-fi/ieee-802-11-standards-tutorial.%20php%20Accessed%2003%20Jan%202015http://www.radioelectronics.com/info%20%20/wireless/wi-fi/ieee-802-11-standards-tutorial.%20php%20Accessed%2003%20Jan%202015http://www.radioelectronics.com/info%20%20/wireless/wi-fi/ieee-802-11-standards-tutorial.%20php%20Accessed%2003%20Jan%202015http://localhost/var/www/apps/conversion/tmp/scratch_2/wseas/Ian%20Poolehttp://www.ncbi.nlm.nih.gov/pubmed/7649149http://www.ncbi.nlm.nih.gov/pubmed/7649149http://www.ncbi.nlm.nih.gov/pubmed/7649149http://www.ncbi.nlm.nih.gov/pubmed/7649149